Statistical advice to authors

In environmental studies a holistic approach to data analysis should be adopted. Environmental parameters (such as temperature, nitrogen, water depth...) should be seen as acting together, and analysed accordingly. Studies in which effects of each environmental parameter are analysed separately might not be considered.

Statistical methods should be reported as thoroughly as other aspects of the methodology, i.e., so that the reader is able replicate the work. If you are not a statistician, consult one, or a good textbook (e.g. Biostatistical Analysis by Jerrold H. Zar).

When using statistical methods, consider the following:

1. The rationale behind the testing should always be presented and comparisons clearly identified.
2. Statistics used should be named and applicability of each test should be given.
3. It should be proven that the data can be tested with the tests used (e.g. normality should be tested).
4. If some data are excluded from the analyses, indicate it clearly and explain why.
5. Sample sizes (n) and their calculation should always be given.
6. Number after a ‘±’ sign should be identified (e.g. SE, SD or CI).
7. Merely labelling differences as significant (including such labels as < 0.05, *, etc.) without indicating test results (test value) and sample sizes (or degrees of freedom) is not acceptable.
8. p values should be presented to a reasonable number of decimal places (usually three or four); true values rather than thresholds (0.05, 0.01, 0.001) should be given.
9. Please note that statistical significance does not indicate the strength of a correlation. For example, achieving a value of p = 0.001 does not mean that the relationship is stronger than if you achieved a value of p = 0.04.
10. There is often little value in giving a percentage instead of the numbers that generate it, especially when the denominator is small.
11. If complex statistical models are being employed, some justification of the model choice may be required, including information regarding the model fit of this and other credible models. This may not be appropriate for the paper (except perhaps in summary), but additional material may be usefully submitted.
12. No research study can ever be perfect, because every design has to balance competing demands and make compromises. The paper should discuss potential weaknesses and or biases in the design, conduct and analysis.